End of Megapixel wars and EF-S buying decisions

Started Feb 23, 2006 | Discussions
David Martin Forum Pro • Posts: 17,056
End of Megapixel wars and EF-S buying decisions

Those of you who are happy with around 8MP will not have to concern yourself in any way with this information, but some of you may be buying EF-S and thinking that in future that the number of pixels in the 1.6 format will increase quite a bit, as it has done historically in the past.

This now appearsrather less likely, and I will put into this post the various bits of information which tend to that conclusion so that you can make your own minds up:

A little while ago Chuck Westfall (Canon Technical) did an interview in which he said that we had come about as far as we could in putting more pixels in - he was actually talking about compact cameras, and I can't remember if he entered the caveat 'at present technical levels'

Unfortunately the post is now been pulled, but I include a link to the site here which confirms the header and gives a link where those who want to check further can do so:

http://news.com.com/Pixel+counting+joins+film+in+obsolete+bin/2100-1041_3-6034570.html

This has now been given further weight by a new interview with Chuck at IR where he talks about the reason for the 30D having the same number of pixels as the 20D - and this one has got the caveat 'at current technical levels:
http://www.imaging-resource.com/NEWS/1140633307.html

Paul Pope, whom some of you may not know has excellent contacts with Canon, and frequently has access to models and prototypes before release, confirms that there was a higher megapixel count 1.6 around which dissappeared - perhaps it is not unreasonable to infer from Chuck's interview that Canon could not get the high MP count to give decent results, although that is not cast-iron, of course:
http://forums.dpreview.com/forums/read.asp?forum=1032&message=17299641

Finally, here are a couple of links to more techinical info, firstly one explaining condsiderations of sensor size and pixel density:
http://forums.dpreview.com/forums/read.asp?forum=1032&message=15240525

And one on diffraction - what it is, and at what pixel size it starts to become important:
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

What this one boils down to is that as you start to go to greater pixel density than the 20D/30D, diffraction really starts to kick in and hurt.
So to summarise:

SO FAR AS WE KNOW AT THE PRESENT TIME (Sorry to shout, but I really thought that needed emphasis),

Pixel count for the respective size sensors given the noise performance Canon likes to give us, yields around:
1.6: 8MP

1.3: 12MP (I know, it seems a bit unlikely that we will see too much of that senssor in future!)
1.0: 22MP

PS - for anyone who is thinking that the Nikon D2000 is APS-C sized, and has 10MP two considerations must be bourne in mind:

the Nikon sensor is 1.5, the Canon 1.6, so in fact to have the same density as the 30D the Nikon would have around 9MP anyway.

The D200 has rather worse noise performance than the 20D - different sensor tehnology, a manufacturer which is not so experienced in NR, - but even so, worse than the 20D.

Things may change, of course, but the question I would now be asking myself before buying EF-S glass is 'Willl 8MP suit me for the foreseeable future?'

Rather than'I will probably get the extra rez I could really do with in the next generation of camera'
That is something of a paradigm shift
--
Regards,
DaveMart

Please see profile for equipment

Peter G Veteran Member • Posts: 5,944
Re: End of Megapixel wars and EF-S buying decisions

David Martin wrote:

Pixel count for the respective size sensors given the noise
performance Canon likes to give us, yields around:
1.6: 8MP
1.3: 12MP (I know, it seems a bit unlikely that we will see too
much of that senssor in future!)
1.0: 22Mp

I was playing around with the same and I got in addition to the 8.2MP at 1.6:

1.5: 9.3MP
1.3: 13.3MP
1.0: 20.9MP.

I dont' mind Canons theory about keeping to the smaller pixels unless they get a breakthrough on performance. But how about more area?

I have always been kind of Annoyed by Canons 1.6 crop when everyone else is aiming at 1.5 including 3rd party lens makers. It would also give a bit more width to a lens like the 17-55. Going for 27.2 to 25.5. I would gladly pay the extra $10 for the slight extra pixel count and slightly wider FOV in APS-C.

What I would really want though is the 12+ MP 1.3 crop 5d. Should be around $2k or less, nice hole in the lineup right there. It would be a kick butt combo.

chris_ah1 Regular Member • Posts: 155
Re: End of Megapixel wars and EF-S buying decisions

I prefer 1.6x crop tbh - whilst it does make things worse at the wide end, it probably does make average glass look a lot better with the reduced image circle needed.

I will however go FF eventually though and I wonder whether my FF lenses will still look good though.

Still, I'm sure there will be another sensor coming soon that will greatly improve on current technology - there always is.

Hopefully I won't have a reason to change body for 3years

Peter G wrote:

David Martin wrote:

Pixel count for the respective size sensors given the noise
performance Canon likes to give us, yields around:
1.6: 8MP
1.3: 12MP (I know, it seems a bit unlikely that we will see too
much of that senssor in future!)
1.0: 22Mp

I was playing around with the same and I got in addition to the
8.2MP at 1.6:

1.5: 9.3MP
1.3: 13.3MP
1.0: 20.9MP.

I dont' mind Canons theory about keeping to the smaller pixels
unless they get a breakthrough on performance. But how about more
area?

I have always been kind of Annoyed by Canons 1.6 crop when everyone
else is aiming at 1.5 including 3rd party lens makers. It would
also give a bit more width to a lens like the 17-55. Going for 27.2
to 25.5. I would gladly pay the extra $10 for the slight extra
pixel count and slightly wider FOV in APS-C.

What I would really want though is the 12+ MP 1.3 crop 5d. Should
be around $2k or less, nice hole in the lineup right there. It
would be a kick butt combo.

PicOne
PicOne Veteran Member • Posts: 6,932
EF-S lenses.. anyone know max coverage?

Anybody know if an EF-S lens could cover a bigger sensor -- ie. could one of these perhaps cover a 1.5 or 1.4 crop sensor size?

OP David Martin Forum Pro • Posts: 17,056
Re: End of Megapixel wars and EF-S buying decisions

chris_ah1 wrote:

I prefer 1.6x crop tbh - whilst it does make things worse at the
wide end, it probably does make average glass look a lot better
with the reduced image circle needed.

I will however go FF eventually though and I wonder whether my FF
lenses will still look good though.

Here is a link with lens info where you can check how your glass does at FF:
http://www.the-digital-picture.com/Default.aspx

Still, I'm sure there will be another sensor coming soon that will
greatly improve on current technology - there always is.

Hmmm...well, I did say here is the data for everyone to make their minds up from!

The whole point of the post though, I should perhaps point out, is that we can't perhaps be quite so confident as we used to be that there will be new sensor with greatly improved technology cooming along soon as we once were!

Hopefully I won't have a reason to change body for 3years

Peter G wrote:

David Martin wrote:

Pixel count for the respective size sensors given the noise
performance Canon likes to give us, yields around:
1.6: 8MP
1.3: 12MP (I know, it seems a bit unlikely that we will see too
much of that senssor in future!)
1.0: 22Mp

I was playing around with the same and I got in addition to the
8.2MP at 1.6:

1.5: 9.3MP
1.3: 13.3MP
1.0: 20.9MP.

I dont' mind Canons theory about keeping to the smaller pixels
unless they get a breakthrough on performance. But how about more
area?

I have always been kind of Annoyed by Canons 1.6 crop when everyone
else is aiming at 1.5 including 3rd party lens makers. It would
also give a bit more width to a lens like the 17-55. Going for 27.2
to 25.5. I would gladly pay the extra $10 for the slight extra
pixel count and slightly wider FOV in APS-C.

What I would really want though is the 12+ MP 1.3 crop 5d. Should
be around $2k or less, nice hole in the lineup right there. It
would be a kick butt combo.

-- hide signature --

Regards,
DaveMart

Please see profile for equipment

OP David Martin Forum Pro • Posts: 17,056
1.3 crop doubtful after release of the 17-55EF-S IS

Peter G wrote:

David Martin wrote:

Pixel count for the respective size sensors given the noise
performance Canon likes to give us, yields around:
1.6: 8MP
1.3: 12MP (I know, it seems a bit unlikely that we will see too
much of that senssor in future!)
1.0: 22Mp

I was playing around with the same and I got in addition to the
8.2MP at 1.6:

1.5: 9.3MP
1.3: 13.3MP
1.0: 20.9MP.

I dont' mind Canons theory about keeping to the smaller pixels
unless they get a breakthrough on performance. But how about more
area?

I have always been kind of Annoyed by Canons 1.6 crop when everyone
else is aiming at 1.5 including 3rd party lens makers. It would
also give a bit more width to a lens like the 17-55. Going for 27.2
to 25.5. I would gladly pay the extra $10 for the slight extra
pixel count and slightly wider FOV in APS-C.

What I would really want though is the 12+ MP 1.3 crop 5d. Should
be around $2k or less, nice hole in the lineup right there. It
would be a kick butt combo.

Me too, Peter - but it looks as though Canon are determined to make us fork out $3k for the 5D - the new EF-S lens at the price point it has come in at seems to make a 1.3 fairly unlikely

-- hide signature --

Regards,
DaveMart

Please see profile for equipment

Peter G Veteran Member • Posts: 5,944
Re: 1.3 crop doubtful after release of the 17-55EF-S IS

David Martin wrote:

Me too, Peter - but it looks as though Canon are determined to make
us fork out $3k for the 5D - the new EF-S lens at the price point
it has come in at seems to make a 1.3 fairly unlikely

I dont' know if lenses tell us anything. 1.6 is here to stay regardless of what lenses are built simply because of the need to compete on price on the low end.

But maybe the 24-104L was designed with both 1.0 and 1.3 in mind. On 1.3 it becomes 29-136, which is none to shabby a walk around general lens.

I don't think Canon will ever say 1.3 it is dead. Because at some point they might see a new competetive advantage to a midsize sensor.

But I think the only two sizes guaranteed are 1.6 and FF.

MitchAlsup Veteran Member • Posts: 4,971
Re: End of Megapixel wars and EF-S buying decisions

David Martin wrote:

Paul Pope, whom some of you may not know has excellent contacts
with Canon, and frequently has access to models and prototypes
before release, confirms that there was a higher megapixel count
1.6 around which dissappeared - perhaps it is not unreasonable to
infer from Chuck's interview that Canon could not get the high MP
count to give decent results, although that is not cast-iron, of
course:
http://forums.dpreview.com/forums/read.asp?forum=1032&message=17299641

I suspect that there was a higher pixel count sensor developed for the 30D. After testing Canon decided that the 20D sensor had better noise performance and/or that the new sensor pushed too many lenses into substandard sharpness. Thus the IQ by some Canon metric degraded with more pixels.

I further suspect that this sensor was plug replacible with the 20D sensor so that Canono could evaluate both and choose the sesor for the 30D at the last possible moment before comitting the 30D BOM into production.

Finally, I suspect that given a choice between a 30D with a 20D sensor and a 30D with a (lets just say) 11 MP sensor that about 50% of us would want the 8MP sensor and another 50% would want the 11 MP sensor.

So the question becomes, have we reached the point where making one IQ parameter 'better' almost necessarily makes another important metric 'worse'?

This is not actually very surprising considering that many here rail for more dynamic range over resolution more-or-less all the time.
--
Mitch

UziY Contributing Member • Posts: 736
By that logic, can't but any EF zoom either

If you want a lens compatible with a future 22MP FF, you need edge to edge sharpness compatible with that sensor. There is not a single zoom lens that fits the bill, and few very primes (obviously all L).
--
Uzi
http://www.pbase.com/uyoeli

OP David Martin Forum Pro • Posts: 17,056
Re: By that logic, can't but any EF zoom either

I'm not saying you should not buy a EF-S lens - where did you get that idea?-just that if you do buy one thinking that the pixel count in the 1.6 is likely to rise a lot, you may be disappointed.

You are quite right of course that 22MP may cause difficulties for a lot of lenses - as CW remarks Canon need to do something about that - but edge to edge sharness is not the end of the world for a lot of uses - some portrait work, for instance.

UziY wrote:

If you want a lens compatible with a future 22MP FF, you need edge
to edge sharpness compatible with that sensor. There is not a
single zoom lens that fits the bill, and few very primes (obviously
all L).
--
Uzi
http://www.pbase.com/uyoeli

-- hide signature --

Regards,
DaveMart

Please see profile for equipment

Guitarman Senior Member • Posts: 1,214
Q: diffraction versus noise limits

your points are well taken.... pretty much the story I think...

and I do believe that there was a bigger sensor that didn't make it because of diffraction...

my question - my fav subject - is with the same size sensor/pixel pitch.. have they got..could they deliver ...another stop of noise..PERFORMANCE
on a 40D...for example...

I believe the pixel war has slowed to a near stop ..because of your points..but what about (canon's cornered market) noise wars?

I would have bought another stop of noise reduction...
no wait...noise performance...INSTANTLY... OVER more pixels..

I have enough pixels... now that we have hit (softly) a wall...
but I have too much noise...still....

I want to get the 85 f1.2 to do some REAL WORK

what do you think?

TOM

OP David Martin Forum Pro • Posts: 17,056
Re: Q: diffraction versus noise limits

Well, that is not really to do with diffraction, Tom, which is a separate subject and a problem which only creeps in gradually at certain Fstops as you get smaller pixels.
The answer to your question is basically no, I am afraid.

The thing is, diffraction aside, the thing that stops you going to ever smaller pixels is noise - that is why Canon did not try to put more pixels in the 30D.

If they could leave the pixel count the same, and reduce noise, they could just as well put in more pixels and keep the noise the same.

So if this is just about the end of the pixel count race, it is also the end of great improvements in less noise.
The 5D will probaly be about as good as it gets for some time.

That is with Chuck's proviso though - barring fundmental technical improvements - but there is no sign of them at the moment - Sony also had to go to a larger sensor size in the R1, when they would have loved to built something smaller, I would have thought.

Guitarman wrote:

your points are well taken.... pretty much the story I think...
and I do believe that there was a bigger sensor that didn't make it
because of diffraction...

my question - my fav subject - is with the same size sensor/pixel
pitch.. have they got..could they deliver ...another stop of
noise..PERFORMANCE
on a 40D...for example...

I believe the pixel war has slowed to a near stop ..because of your
points..but what about (canon's cornered market) noise wars?

I would have bought another stop of noise reduction...
no wait...noise performance...INSTANTLY... OVER more pixels..

I have enough pixels... now that we have hit (softly) a wall...
but I have too much noise...still....

I want to get the 85 f1.2 to do some REAL WORK

what do you think?

TOM

-- hide signature --

Regards,
DaveMart

Please see profile for equipment

dbuhanan Contributing Member • Posts: 875
Excellent post David. Makes sense to me... (nt)
-- hide signature --

kindest regards
Dale

It is possible to make a small fortune in photography, you just have to start with a large fortune.

Guitarman Senior Member • Posts: 1,214
Re: Q: diffraction versus noise limits

OK...
but my question goes to:

keep the same size pixels.. pitch ...etc
take the 20D/30D sensor...
now squeeze better noise performance out of it..

what are the limits..? shorter paths? better paths?

obviously we are at the propriitary basic research level
of canon...here...
but I believe they CAN get another stop.. better noise..

if they STOP trying for more pixels.. and 'tweak' the processes
for noise performance..

yeah - I know ...wishfull thinking..
but there are always some efficiencies.. lying on the table..

I am talking WITHOUT cryo cooling....

plus use
lower noise amps ??.... outboard etc...

hey ...give me hope... otherwise...
I might have to go out and take a picture...ha!

anyway, thanks for the useful topic...
I am trying to get my girlfriend to use the 20D more .. so I can whine..
and buy a 30D...

me ...I LIKE 1.6 crop...

maybe they are holding-out to put that 6400 iso into a FF$8000 camera... I will get one if they do..
i.e. ....if the 1600 is as clean as a 20D 400/800

Later
TOM

OP David Martin Forum Pro • Posts: 17,056
I want ' The war to end wars' here - about megapixel density!

So in an associated thread here someone had some good arguments - so let's get it on guys!
I referred him to this thread - hopefully he willl post here.

dbuhanan wrote:

-- hide signature --

Regards,
DaveMart

Please see profile for equipment

Jeff Varszegi Senior Member • Posts: 2,967
Questions

Putting diffraction aside for a minute, assuming that there are potential methods for minimizing it or doing away with it entirely:

If increasing pixel count means a degradation in noise performance, couldn't it (in a perfect world) be countered by the increased pixel count? Meaning: if you doubled the pixel count on both the X and Y axis (quadrupling the number of pixels), then took a noisy low-light picture, then resized the image to one-quarter its size, would the noise be about the same as a picture from the original sensor design?

Also, if the performance thus gotten could be roughly equal, wouldn't you want the extra effective resolution in good light, when noise isn't an issue?

thx1138
thx1138 Veteran Member • Posts: 7,162
Re: Questions

Jeff Varszegi wrote:

Putting diffraction aside for a minute, assuming that there are
potential methods for minimizing it or doing away with it entirely:

If increasing pixel count means a degradation in noise performance,
couldn't it (in a perfect world) be countered by the increased
pixel count? Meaning: if you doubled the pixel count on both the X
and Y axis (quadrupling the number of pixels), then took a noisy
low-light picture, then resized the image to one-quarter its size,
would the noise be about the same as a picture from the original
sensor design?

Also, if the performance thus gotten could be roughly equal,
wouldn't you want the extra effective resolution in good light,
when noise isn't an issue?

Yes the noise will be significantly improved by downszing, however, wouild you acept the loss in detail? All interpolation schemes are a trade off and you cannot reproduce the same detail in an uprezzed or downrezzed image. Sharpening does not help. But maybe the trade off is worth it for those low light cases as super fine detail might not be an issue anyway.

Personally I find Canon's arguments suspicious. Look at how well the 20D does with 6.4um pixels compared to the 5D/1D2 8.2um pixels or 1Ds2 7.2 um pixels. There's nothing inherent about 6.4um pixels in the physics that makes that some special cutoff. I find it hard to believe Canon jumped immediately to this size pixel with the 20D thinking, well that's it we can't do better. Fuji certainly doesn't think so and looking at their f10 with it's order of magnitude better noise than competitors there seems to be some method to their madness. I wonder what Canon is thinking of to address low light noise performance. The 1.6x crop sensor will not stay at 8Mp forever. It could be the case shortly that Canon having led the Mp race for so long will be the only manufacturer with less than 10Mp within 12 months. Also interesting to seeis what Fuji's S4 pro will be like. I have a feeling based on what they said a few months ago, this thing will have an interpolated 18-22Mp, but with class leading noise.

Anyway, whether or not the 1.6x crop sensor has currently hit the wall in terms of Canons standards for image quality, the fact is they could easily scale this pixel pitch to the 1.3x crops sensor and give us a 13.3Mp sensor with exactly the same perfomance, and I know which sensor I'd rather have in my camera. Those extra 5MP will come in very handy when for landscape/macro/portrait and printing. Look at the detail the 5D is getting from 12.8MP.

Eric Senior Member • Posts: 1,781
Great post -- Here are my three points:

Bringing these over from another thread by request:

1) If photozone.de's tests are of any value at all, it's clear that only a select few lenses can resolve beyond 8.2mpx, and even then only at a select few apertures and focal lengths. What is gained with 10mpx?

2) Cramming 10+mpx on the 30D would have increased both noise and price, correct? Since the 30D was not intended to appeal as a potential upgrade to 20D owners as much as 300/350D owners, wouldn't a larger price gap and more noise hurt sales? Most educated buyers are over the whole "the more pixels the better" syndrome by now, correct?

3) Aside from all of that, I'd also wonder how often people are going to be printing prints large enough to benefit from 10mpx over 8mpx. (That's assuming Canon makes one with equal or less noise, and you have a lens that can utilize such a small pixel pitch). How often does an amateur/hobbyist photographer print larger than 12x16? Unless you're making huge prints of severe crops, how often will you notice the difference?

  1. 3 is more of a theoretical "what's it worth" point. #1 and #2 are technical and I don't know if they've even been refuted yet.

Thanks!

  • Eric

Ian Hobday Regular Member • Posts: 455
Good post, one thing to add...

Very well said.

One additional thing worth noting is that you need a substantial increase in the pixel count before you see a worthwhile increase in horizontal or vertical resolution.

10D: 3072 x 2048
20D: 3504 x 2336
D200: 3872 x 2592
5D: 4368 x 2912

10D to 20D:
Pixel count increase: 30%
Horz res inc: 14%
Vert res inc: 14%

20D to D200:
Pixel count increase: 22%
Horz res inc: 11%
Vert res inc: 11%

20D to 5D:
Pixel count increase: 55%
Horz res inc: 25%
Vert res inc: 25%

In my opinion a 25% increase in horz/vert resolution is worthwhile. Less than that and the resolution increase on its own is not worth the upgrade -- there would need to be other very important improvements to make it worthwhile.

Ian

Guitarman Senior Member • Posts: 1,214
Re: Great post -- Here are my three points:

I was SURE they were gonna reduce noise by one stop..
on this new camera...
and I didn't see the need for the increased pixels..

the body-blow to NIKON would have been even better noise performance...
but they didn't do it !!!???

Now I think they are gonna leave it there for a while..
and MAYBE improve noise on the more expensive cameras..

does anyone think there is more noise performance to be had.. with the
20D/30D sensor config .... pixel pitch?

Keyboard shortcuts:
FForum MMy threads